Modeling Cue-Based Retrieval and Prediction – One Morpheme at a Time
Abstract
This work proposes an extension to the cue-based retrieval theory of sentence processing: memory retrieval and predic- tion processes during sentence comprehension take place at the morpheme level instead of at the word level. We illustrate this proposal by extending an existing cue-based retrieval model from word-level to morpheme-level processing, and show that our model better captures the interactions between memory re- trieval and predictive processing. Specifically, we extend the model reported in Patil and Lago (2021), which accounted for the interaction between retrieval and prediction during the pro- cessing of German possessive pronouns, but failed to general- ize to structures involving determiners (Oltrogge, Ver´ıssimo, Patil, & Lago, accepted). Our results show that modeling at the morpheme level captures retrieval-prediction interactions more precisely. The model successfully predicts the pattern of prediction onsets across German possessive pronouns and determiners. The proposed morpheme-by-morpheme model is further supported by psycholinguistic evidence suggesting that humans naturally decompose words into their constituent mor- phemes